Pinned_allocator.h not available in CUDA 12.6

[Andymulb]

Hi, thanks for your excellent work!

I'm trying to run the system on resource-constrained hardware (NVIDIA Jetson AGX Orin). I already tested various differentiable or partially differentiable SLAM systems. However, the runtime wasn't anywhere near real-time performance. I found your paper a while ago but couldn't install X-SLAM on the AGX Orin. This week, I tried it again and I receive this error message when trying to build X-SLAM:

[  5%] Building CXX object DeviceArray/CMakeFiles/device_array.dir/src/DoubleComplex.cpp.o
[ 10%] Building CXX object DeviceArray/CMakeFiles/device_array.dir/src/device_memory.cpp.o
In file included from /home/X-SLAM/DeviceArray/src/device_memory.cpp:5:
/home/X-SLAM/Common/include/cx.h:29:10: fatal error: thrust/system/cuda/experimental/pinned_allocator.h: No such file or directory
   29 | #include "thrust/system/cuda/experimental/pinned_allocator.h"
      |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
compilation terminated.
make[2]: *** [DeviceArray/CMakeFiles/device_array.dir/build.make:93: DeviceArray/CMakeFiles/device_array.dir/src/device_memory.cpp.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:221: DeviceArray/CMakeFiles/device_array.dir/all] Error 2
make: *** [Makefile:91: all] Error 2

Apparently, the header file "pinned_allocator.h" is no longer available in CUDA > 12. Do you know how I can run the code on the AGX Orin (arch 87) with CUDA 12.6?